I just noticed that we insert a node kind into a 'base-deleted' row, at least 
in some cases.

wc_db.c: svn_wc__db_extend_parent_delete(... kind ...):  stmt = 
STMT_INSTALL_WORKING_NODE_FOR_DELETE

  svn_sqlite__bindf(... kind ...)

Is there a good reason for this?  It makes me feel that it is redundant and 
therefore likely to get forgotton, get out of sync and lead to hard-to-discover 
errors.

Does anyone have an objection to me making the rule be that the 'kind' field 
should be left as null when presence==base-deleted, and figuring out what fixes 
we need to make to the code to implement that?

- Julian

Reply via email to